Crystal facet engineering of metal oxides for upgrading biomass-derived oxygenates: a perspective

Abstract

Coordinatively unsaturated metal cations and oxygen anions on the surface of amphoteric metal oxides serve as acid and base centers, respectively, enabling a number of acid/base-catalyzed reactions. In particular, the O atom of oxygenates can readily coordinate with surface metal cations and can therefore initiate various reactions that are important for upgrading biomass-derived oxygenates. Metal oxide catalysts prepared via conventional methods expose various facets with distinct properties, making it difficult to elucidate the reaction chemistry and mechanism mediated on metal oxides at an atomic level. Advances in synthesis of nanocrystals of metal oxides with predominantly exposing a certain facet allow one to explore the reaction chemistry on well-defined metal oxide facets under real reaction conditions and thus may provide a powerful approach to investigate the facet-dependent structure–activity relationships of metal oxide-mediated reactions. In this perspective, we first elaborated on the fine structure and properties of various facets of TiO₂ and CeO₂ and then surveyed the dehydration/dehydrogenation of alcohols, ketonization of carboxylic acids, and aldol condensation of ketones and aldehydes mediated by metal oxides. In particular, we discussed the relationship between catalytic performances and the facet-dependent acid–base property and fine surface geometrical structure. Finally, we described challenges and opportunities associated with future research on metal oxide-mediated reactions via crystal facet engineering.

---

1. Introduction

Metal oxides have been widely used as supports and even catalysts for catalyzing various reactions.^1–4 In particular, amphoteric oxides, such as TiO₂, ZrO₂, and CeO₂, have been widely explored owing to their mild acid–base properties.^5,6 On the surface of such metal oxides, coordinatively unsaturated metal cations can accept electrons and thus act as Lewis acid centers to stabilize reactant ions. In contrast, coordinatively unsaturated O anions act as Brønsted base centers, which can abstract H from hydrocarbons and stabilize protons. In particular, the synergy of the acid site and adjacent base site, i.e., an acid–base pair, have been reported to be able to heterolytically dissociate H₂, C–H, and O–H bonds of various molecules (Scheme 1)⁷ and thus mediate various organic reactions.^7–10

Scheme 1 Dissociative adsorption of (a) hydrogen, (b) methane, (c) water, (d) carboxylic acids in bidentate and monodentate configurations and (e) alcohols on the acid–base pair of the metal oxide surface. M represents the metal, and R represents the alkyl group.

The strength of the acid and base sites on a metal oxide surface is largely dependent on the coordination state of the metal cation and oxygen anion, respectively. In general, lower the coordination number of a metal cation and oxygen anion, higher the ability to accept electrons and donate electrons and, therefore, stronger the acidity and basicity, respectively. In addition, the densities of acid and base sites are strongly dependent on surface configuration. Hence, both the density and strength of an acid–base pair are determined on the basis of the exposed facet of metal oxide.

In addition to the acid–base property, the catalytic performance of a metal oxide depends on the geometric configuration of acid and base sites, the presence of adsorbed H₂O, and the presence of oxygen vacancies. For a certain metal oxide, these properties are largely determined by the exposed facets. However, metal oxide surfaces are complex and may expose various facets with different fractions. The distinct surface properties between various facets and the complex surface composition of facets make it difficult to identify the most active facet as well as to establish rigorous structure–activity relationships at an atomic level.

Investigating reactions using metal oxide catalysts with well-defined surface structures can certainly help understand the reaction chemistry mediated by acid–base pairs on metal oxides. To this end, the surface science approach using a single crystal of metal oxide with a controlled exposed facet has been extensively applied to study the reaction mechanism under high vacuum conditions.^11–14 Although significant advances have been made, the pressure gap (high vacuum vs. atmospheric or high pressure) and material gap (extended surface vs. nanoparticle) between surface science and real reaction conditions may limit the application of the conclusions derived from the surface science experiments. On the other hand, the recent advances in the synthesis of nanocrystal catalysts with well-controlled exposed facets make the investigation of reactions under real conditions possible, and hence may bridge the “pressure gap” and “material gap”. Moreover, engineering the nanocrystals with well-controlled facets via advanced synthesis approaches allows one to investigate the reactions far beyond high vacuum conditions and can be extended to thermocatalysis, electrocatalysis and photocatalysis. Therefore, crystal-facet engineering provides a powerful approach to investigate the structure–activity relationship of reaction chemistry mediated by metal oxides at the atomic level.

As an alternative and/or supplementary resource to fossil fuels, the conversion of renewable biomass to sustainable fuels and chemicals plays an increasing role in the development of our society. The lignocellulose biomass can be decomposed to oxygenated hydrocarbons through pyrolysis, hydrolysis, liquefaction, and fermentation processes (Scheme 2).^15,16 These oxygenates contain hydroxyl (–OH), carbonyl (–CO–), carboxyl (–COOH) and other functional groups, which result in high oxygen content, low heating value, low stability and high corrosiveness. In particular, the cellulose and hemicellulose of biomass can be decomposed to short-chain oxygenates, such as short-chain alcohols, ketones/aldehydes and carboxylic acids. Animal and vegetable fats and oils, as an important biomass resource, can be hydrolyzed to glycerol and long-chain carboxylic acids. In addition, fermentation of wet waste can produce C₂–C₈ volatile carboxylic acids.^17,18 Thus, upgrading these oxygenates (alcohols, ketones/aldehydes, and carboxylic acids) into more stable and valuable fuels and chemicals is desirable. Indeed, the amphoteric metal oxides have been widely applied in these upgrading processes through catalyzing various organic reactions, such as dehydration, aldol condensation, ketonization and others,^19–21 which produce valuable chemical products of olefins, longer chain ketones and aldehydes (Scheme 2). During these reactions, the oxygen atom of oxygenates readily coordinates to the surface coordinatively unsaturated metal cation (Lewis acid site) of the metal oxide, stabilizing the reactants and intermediates. In the meantime, the H atom from either the –OH group or C–H of the oxygenate reactant may be transferred to the surface coordinatively unsaturated O atom (base site) of the metal oxide to form a reactive intermediate and initiate the reaction. These reactions may strongly depend on the acid–base properties and the surface configurations of the metal oxide. Hence, this perspective focuses on the recent advances in crystal-facet engineering for the conversion of biomass-derived oxygenates on TiO₂ and CeO₂ metal oxides.

Scheme 2 The origin of biomass-derived oxygenates (alcohols, carboxylic acids, and ketones/aldehydes) and their conversion into chemical products catalyzed by metal oxides.

A number of reviews relating to the topic are available in the literature.^22–25 These reviews mainly focused on the synthesis methods of metal oxides with different facets and reactions catalyzed by different facets of metal oxides. However, the analysis and summary of the structure–activity relationship between the metal oxide facet and catalytic performance at an atomic level are rare. In this perspective, we intend to summarize and analyze the advances in crystal-facet engineering for the conversion of biomass-derived oxygenates (alcohols, carboxylic acids and aldehydes/ketones) on different crystal facets of TiO₂ and CeO₂, focusing on the structure–activity relationship at the atomic level. We begin with the coordination environments and configuration structures of distinct crystal facets. Then, we discuss the structure–activity relationship between the crystal facet of metal oxide and the catalytic performance of biomass-derived oxygenates. Finally, we provide future challenges and opportunities in crystal-facet engineering.

2. Synthesis and structure of different crystal facets of TiO₂ and CeO₂

The synthesis of metal oxides exposing certain facets and understanding the fine surface structure of different facets are the keys to crystal-facet engineering. In addition, correlating the fine structure of crystal facets with physical and chemical properties is also important to understand the reaction chemistry mediated on metal oxides. In this section, we briefly summarize the synthesis of TiO₂ and CeO₂ with controlled exposed facets, as well as the fine structure of a specific facet.

2.1. TiO₂

TiO₂ has three distinct crystal phases, anatase, rutile and brookite, among which anatase is the most widely used.²⁶ Anatase TiO₂ nanocrystals with a well-defined multi-faceted structure can be synthesized by a variety of methods, including hydrothermal, solvothermal, gas oxidation and amorphous TiO₂ crystallographic transformation.²⁷ Among these methods, the hydrothermal and solvothermal methods using F⁻, SO₄²⁻, and Cl⁻ anions as additives were widely used for the synthesis of (001), (100), and (101) predominant facets, respectively, due to the moderate synthesis condition and simple procedure.^28–30 Additionally, utilizing urea as a capping agent to prepare TiO₂ with a predominant (001) facet has also been explored,⁹ which offers an approach with few impurities being introduced to the crystal since urea is readily decomposed.

In the bulk anatase TiO₂, the O and Ti atoms (O_3c and Ti_6c) are coordinately saturated with three Ti and six O atoms, respectively. In contrast, the surface O and Ti atoms are typically coordinatively unsaturated. The fine coordination and configuration structures vary significantly on different facets (Fig. 1). The TiO₂(101) facet shows a sawtooth-like structure with O_2c in the first layer and Ti_5c and O_3c in the second layer. The shortest distances between Ti_5c–Ti_5c and Ti_5c–O_2c are 3.86 and 1.85 Å, respectively. The TiO₂(100) facet exhibits a stepped structure with O_2c, O_3c, and Ti_5c in the first layer. While the Ti_5c–O_2c distance is similar to that of TiO₂(101) at 1.86 Å, the Ti_5c–Ti_5c distance is much shorter, i.e., 3.08 Å. The TiO₂(001) facet shows a nearly flat surface, with O_2c, Ti_5c and O_3c in the first, second and third layers, respectively. The Ti_5c–O_2c–Ti_5c angle is 153°, the shortest distance of Ti_5c–Ti_5c is 3.86 Å while the shortest distance of Ti_5c–O_2c is longer (1.99 Å) compared to other facets. Such a unique surface configuration places stress on the Ti_5c–O_2c bond and makes the O_2c very reactive. These varied surface configurations result in the density of Ti_5c–O_2c pair on TiO₂(101), (100) and (001) facets being 5.08, 5.49, and 6.70 nm⁻², respectively, which imply the (001) surface has the highest density of the acid–base pair in terms of surface area.

Fig. 1 Surface configuration of the (101), (100), and (001) facets of TiO₂. Light blue and red spheres are Ti and O atoms, respectively. The distances in the figure are given in Å.³¹

The estimated surface energies are 0.77, 0.87 and 1.41 J m⁻² for TiO₂(101), (100) and (001), respectively, indicating the (001) facet is less stable,³² which is in agreement with the longer Ti_5c–O_2c distance for the (001) facet. Also, the longer Ti_5c–O_2c distance may also make the formation of oxygen vacancies facile. Indeed, the oxygen vacancy formation energies of TiO₂(101), (100) and (001) facets are 3.48, 3.32 and 2.80 eV, respectively,³³ which are in line with the surface energies and point to the lower stability the easier formation of oxygen vacancies. The density of surface oxygen vacancies will change the surface configuration and influence the acid–base property of adjacent Ti–O pairs, which may eventually affect the reactions mediated on the Ti–O pair. Moreover, varied surface reactivity will also influence the H₂O adsorption when the catalyst is exposed to atmosphere in the presence of H₂O. It has been reported that the reactive Ti_5c–O_2c on TiO₂(001) favors dissociative adsorption of H₂O, while the less reactive pair on the TiO₂(101) facet favors molecular adsorption of H₂O.^34–37 Apparently, the dissociated H₂O occupies the original Ti_5c–O_2c pair, and thus alters the acid–base properties of the occupied Ti–O pair and adjacent unoccupied Ti–O pairs, which eventually would influence the reactions mediated by the acid–base pair.

2.2. CeO₂

The CeO₂ has a fluorite structure, with Ce cations being arranged in the face center and corner of the cubic structure and O anions occupying the tetrahedral interstices of the face-centered cubic Ce ions. The CeO₂ nanocrystals with predominantly exposed low-index facets of (111), (110), and (100) can be prepared by hydrothermal methods. Unlike the synthesis of TiO₂ using different anions, the main factors determining the exposed facet of CeO₂ are the alkali concentration and the hydrothermal temperature.^38–40 Low OH⁻ concentration (0.01 mol L⁻¹) and low temperature (100 °C) prompted the generation of octahedral structures with the dominantly exposed (111) facet. Increasing the OH⁻ concentration to 6 mol L⁻¹ yielded nano-rod structures with (110) as the dominantly exposed facet. Increasing the hydrothermal temperature (180 °C) at high OH⁻ concentrations, cubic structures of CeO₂ were formed with dominantly exposed (100) facets.

The Ce and O atoms in the bulk CeO₂ are coordinatively saturated with 8 O and 4 Ce (Ce_8c and O_4c), respectively. As illustrated in Fig. 2, the (110) facet shows O_3c and Ce_6c atoms in the topmost layer. In comparison, both (111) and (100) facets show a sawtooth-like configuration. The (111) facet contains O_3c and Ce_7c in the first and second layers, respectively. Note that half of the O atoms in the outmost layer were removed from the (100) facet to achieve a stable surface. As a result, the (100) facet shows O_2c and Ce_6c in the first and second layers, respectively.

Fig. 2 Surface configuration of the (110), (111), and (100) facets of CeO₂. Light blue and red spheres are Ce and O atoms in the first layer, respectively. Light blue and brown spheres are Ce and O atoms in the second and third layers, respectively. The distances in the figure are given in Å.⁴¹

Although the bond lengths of Ce–O pairs and the shortest distances of Ce–Ce pairs do not differ much among these facets, the differences in geometrical configuration (Fig. 2) result in a change in the distribution and density of Lewis acid–base pairs. The densities of Ce atoms per nm² surface on (110), (111), and (100) are 4.78, 7.81, and 6.64, respectively, indicating that the (111) facet exhibits the highest density of Lewis acid sites.^42,43 The energies for the formation of oxygen vacancies on CeO₂(110), (111) and (100) facets are determined to be 2.69, 3.30 and 2.97 eV, respectively,⁴⁴ which imply that the (110) facet may have the highest density of oxygen vacancies. The presence of oxygen vacancies may further affect the acid–base properties of different facets.^45–47

2.3. Surface reconstruction

Despite the pristine facet of metal oxides showing well-defined structures, the real facet of metal oxides may undergo reconstruction under certain conditions. This is particularly true for the facet with high surface energy since the reconstruction reduces the surface energy and thus improves the stability of the surface. For example, the anatase TiO₂(001) surface tends to undergo a (1 × 4) reconstruction, particularly under the conditions of high temperature (>773 K) and ultrahigh vacuum or the conditions of ion sputtering or electron-beam irradiation.^48–50 Although various models have been proposed for the (1 × 4) reconstruction, the ad-molecule model (ADM) is now widely accepted, which involves the periodic replacement of rows of surface bridging oxygen on the original (1 × 1) surface with rows of TiO₃ species.^48,51,52 This reconstruction results in the formation of Ti_4c on the ridge, releases the stress on the Ti_5c–O_2c pairs on the surface, and eventually lowers the surface energy from 0.90 to 0.51 J m⁻². Similarly, the CeO₂(110) crystal surface was also reported to undergo a (2 × 1) reconstruction to lower the surface energy during annealing at 940 °C under ultra-high vacuum conditions.⁵³ Reconstructions also occur on other facets of TiO₂ and CeO₂ crystals, such as (1 × 2) reconstruction on the rutile (110) surface,⁵⁴ (1 × 3) reconstruction on the rutile (100) surface,⁵⁵ and the various reconstructions on CeO₂(100).^56,57 It is evident that these reconstructions will change the coordination and configuration of surface fine structures and may significantly influence the adsorption and reaction mediated by the metal oxide surface. Note that the reported reconstruction typically occurs under harsh conditions, i.e., high temperature and high vacuum, which is far from the real reaction conditions for conversion of biomass-derived oxygenates (typically at atmospheric or high pressure and temperature <773 K). Hence, one may anticipate that the surface reconstruction is limited and has little effect on the reaction, which was ignored in most of the previous studies. However, some literature studies have reported that the surface reconstruction takes place under mild conditions.^52,58,59 It is therefore still needed to understand the extent of reconstruction of metal oxide facets under real reaction conditions.

2.4. Crystal facet-dependent properties

As discussed above, the distinct coordination numbers for both metal cations and O anions on different facets can apparently alter the strength of acid/base sites. Also, the varied fine surface geometrical configuration of metal cations and O anions on different facets certainly change the density of acid/base sites. Moreover, the fine surface spatial configuration of metal cations and O anions on different facets may affect the adsorption structure of the reactant as well as the transition state structures along the reaction coordinate. In addition, the varied coordination and configuration structures of the surface O anions on different facets can further affect the density of oxygen vacancies on the metal oxide and hence may affect the adsorption and reaction of reactants.^60–62 Collectively, the acid–base properties, geometrical configurations and density of oxygen vacancies are all dependent on the facets of a metal oxide, which can collectively and/or independently affect the reaction chemistry mediated by metal oxides. In the following sections, we discuss the effect of facet-dependent acid–base properties and geometrical configurations on the metal oxide-mediated conversion of biomass-derived oxygenates.

3. Facet-dependent metal oxide-mediated conversion of oxygenates

Compared to metal oxide catalysts with heterogeneous surface structures, crystal-facet engineering using metal oxides with specific morphologies with varying predominantly exposed facets allows one to more intuitively and accurately determine the surface structure of a catalyst and thus offers a straight and convenient approach to understanding the structure–activity relationships for reactions mediated on metal oxides. For a number of reactions, the varied catalytic performances of varying facets have been widely reported and have been correlated with different facet-dependent properties, such as acid–base properties, geometrical configurations and density of oxygen vacancies. Therefore, in this section, we survey and discuss the influence of facet-dependent properties caused by the effect of morphology on reactions.

3.1. Dehydration of alcohols

The acid–base property is the primary and one of the most important properties of amphoteric metal oxides, which plays an essential role in the adsorption of reactants and stabilization of intermediates and transition states of the reaction through acid–base interactions. Thus, despite other factors, such as surface configuration and oxygen vacancies playing the predominant role in some reactions, the acid–base property should affect all reactions mediated by amphoteric metal oxides.

Dehydration of alcohols to olefins is a typically acid-catalyzed reaction. Two mechanisms, i.e., E1 and E2, have been proposed for dehydration (Fig. 3).^63–65 The E1 reaction mechanism involves the stepwise cleavage of the C–OH and C–H bonds, which can further be classified as E1 (C–OH cleavage first) and E1_cb (C–H cleavage first). Differently, the E2 mechanism assumes that the C–O and C–H bonds are cleaved simultaneously. Metal oxide catalysts have been widely investigated for this reaction^66–68 and were suggested to follow the E2 mechanism, in which Lewis acid sites (metal cations) act as the active centers when paired with vicinal basic surface O atoms or OH groups.^66,69

Fig. 3 E1 and E2 mechanisms involved in the dehydration of isopropanol.⁶⁹

To study the reaction process at the atomic level, Lin et al. studied the dehydration of 2-propanol over anatase TiO₂ with dominant exposure of (101) and (001) facets, respectively.⁷⁰ Kinetic experiments indicated that the (101) facet is more active toward dehydration over the (001) facet, and the former facet shows much lower activation energy than the latter one (131 ± 8 vs. 267 ± 46 kJ mol⁻¹). Temperature programmed desorption (TPD) of NH₃ and diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) of pyridine adsorption results collectively demonstrated that the acid density is higher on the (101) facet (1.14 sites per nm²) than on the (001) facet (0.72 sites per nm²). The TPD of pyridine in IR also indicated the stronger Lewis acidity of the (101) facet than that of the (001) facet. The DRIFTS of 2-propanol adsorption showed the adsorption configuration is strongly dependent on the facet exposed. The (101) facet favors molecular adsorption, while the (001) facet favors dissociative adsorption. Further, theoretical density functional (DFT) calculations of 2-propanol dehydration on hydroxylated (101) and (001) facets via the E2 mechanism (Fig. 4) confirmed that the (101) facet has a lower barrier than that of the (001) facet by 39 kJ mol⁻¹. Similarly, when methanol was employed as a probe molecule,⁹ molecular and dissociative adsorption of methanol were identified on TiO₂(101) and (001), respectively. Interestingly, we noted that the measured higher acid density on TiO₂(101) than that on TiO₂(001)⁷⁰ disagrees with the theoretically predicted bare facets, which show a higher density of Ti_5c on the (001) facet than on the (101) facet. This discrepancy likely results from the presence of H₂O on the surface of catalysts under real conditions. For instance, the (001) facet favors the dissociative adsorption of H₂O,^34–37 which forms –OH groups and occupies the surface Ti–O pairs, resulting in reduced acidity. Indeed, Lin et al. used dissociated H₂O-covered (101) and (001) facets to simulate the dehydration of 2-propanol, and this led to good agreement between the simulated activity and experimental results.⁷⁰

Fig. 4 Potential energy profiles of 2-propanol dehydration on hydroxylated anatase TiO₂(101) and (001).⁷⁰

The conversion of 2-propanol on CeO₂ nano crystals was investigated by Sudduth et al.⁷¹ The CeO₂ nano cubes (CeO₂-C) and octahedrons (CeO₂-O) with a dominant facet of (100) and (111), respectively, were synthesized and used as catalysts. Characterizations showed that the density of the acid site on CeO₂-O is 1.5 times higher than that on CeO₂-C, and the acid strength is also stronger on CeO₂-O. The densities of the basic site on both samples are similar, but the strength of the CeO₂-C is stronger. These differences in acid–base properties affect the conversion of 2-propanol toward different products. The reaction rate of dehydration toward propylene (6 kPa 2-propanol, 270 °C, WHSV = 7.5 h⁻¹) on CeO₂-O is 0.0062 μmol m⁻² s⁻¹, which is much higher than that on CeO₂-C (0.0003 μmol m⁻² s⁻¹), whereas CeO₂-C shows a higher rate of dehydrogenation toward acetone than CeO₂-O (0.0141 vs. 0.0093 μmol m⁻² s⁻¹). These results are consistent with the measured acidity and basicity since the dehydration and dehydrogenation are catalyzed by acid and base sites, respectively. Note that the measured stronger basicity of CeO₂-C agrees well with the lower coordination number of O on the (100) facet, and the higher acid density of CeO₂-O also correlates well with the higher surface dentistry of Ce cations on the (111) facet.

These works clearly indicate that the acid–base property of metal oxides can be tuned by varying the exposed facets. Moreover, the acid–base property can be used to explain the activity and selectivity of the reactions of dehydration and dehydrogenation of alcohols well. One may expect that the unimolecular reaction of dehydration or dehydrogenation only requires one active center, i.e., one acid–base pair, which results in the acid–base property as the predominant factor that affects the activity and selectivity. Note that the measured acid–base property may not always agree with what is theoretically predicted for the pristine facet. Under such circumstances, the presence of surface molecular and/or dissociated H₂O under real reaction conditions may be considered to tune the surface acid–base properties of various facets.

3.2. Ketonization of carboxylic acids

Ketonization reaction converts two carboxylic acids into a ketone with simultaneous elimination of H₂O and CO₂, which increases the carbon chain length and removes 3/4 O without consuming costly H₂. The metal oxide-mediated ketonization initiates with dissociative adsorption of a carboxylic acid at one or two acid–base pairs on the surface (Fig. 5). Subsequently, abstraction of the α-H from the carboxylate by surface basic O anion forms an enolate, which nucleophilically attacks the carboxylic C of another carboxylic acid to form a β-keto acid intermediate and finally decomposes to ketone. The C–C coupling (nucleophilically attack) step is usually believed to be the rate-limiting step rather than the α-H abstraction due to the strong adsorption of carboxylic acid, which makes the α-H abstraction step in quasi-equilibrium.^72,73 In this reaction, the surface metal cations stabilize both carboxylates and the transition state complex, while surface O anions enable α-H abstraction. Hence, this reaction requires both acid and base sites.

Fig. 5 Possible ketonization mechanism through (a) bidentate propionate and (b) monodentate propionate as active species on CeO₂.⁴¹

TiO₂ catalysts with varied crystal phases and facets have been intensively investigated for ketonization of carboxylic acids.^74–77 Wang and Iglesia compared the ketonization of C_2–4 short-chain carboxylic acids on anatase and rutile TiO₂ ⁷² and found that the ketonization turnover rate on anatase is 5 times higher than that on rutile at 523 K and 1.0 kPa acetic acid. IR results showed that bidentate carboxylate is the major surface species on rutile, whereas monodentate carboxylate is the major one on anatase. In comparison with the activity, it suggests monodentate carboxylate is the more reactive intermediate toward ketonization. Their DFT simulation results demonstrated that the Ti–O pair with moderate acid–base strength is the active site and the distinct surface structures of TiO₂(101) and (110) facet (the most stable facets for anatase and rutile, respectively) exert the major surface intermediate. The shorter Ti_5c–Ti_5c distance of TiO₂(110) (0.299 nm) relative to TiO₂(101) (0.383 nm) matches better with the O–O carboxylate distance (0.231 nm) and thus stabilizes the bidentate carboxylate better, resulting in the inactive bidentate carboxylate as the dominant species on rutile while active monodentate carboxylate is the dominant species on anatase. Moreover, the longer Ti_5c–O_2c bond length of the (110) facet than the (101) facet increases the distance between the transition state complex and surface hydroxyl groups, resulting in the absence of hydrogen bonding to stabilize the transition state further. In contrast, the H bonding is present on the (101) facet with a shorter Ti_5c–O_2c bond length. These differences in surface structures make anatase more active than rutile for ketonization.

Recently, Huang et al. studied the ketonization of propionic acid over anatase TiO₂ with dominantly exposed (101), (100) and (001) facets, respectively.³¹ At 350 °C, the complete conversion of propionic acid on TiO₂(001) was achieved at a space time (W/F) of 0.5 h, which is much shorter than 1.2 h for TiO₂(101) and TiO₂(100) (Fig. 6a). The selectivity of 3-pentanone were close 100% on all three catalysts. The turnover frequency (TOF) of ketonization at 350 °C on TiO₂(001) is 185.2 h⁻¹, which is also much higher than that of 71.3 and 51.8 h⁻¹ on TiO₂(101) and TiO₂(100), respectively. The IR band intensity ratio of monodentate carboxylate to bidentate carboxylate decreased with increasing desorption temperature on all TiO₂ catalysts (Fig. 6b), and the decrease rate followed the order TiO₂(001) > TiO₂(101) > TiO₂(100), which suggest the monodentate carboxylate is more reactive than the bidentate one toward ketonization and the TiO₂(001) is more active than other facets. The DFT results of fully propionate-covered TiO₂ surfaces (Fig. 6c) demonstrated that the shorter Ti_5c–Ti_5c distance (3.08 Å) of the (100) facet leads to better stabilization of bidentate propionate and makes it the dominant species. In addition, the shorter distance also imposes strong lateral repulsion between the methyl groups of monodentate propionates (Fig. 6b), which may hinder the formation of a C–C coupling transition state for ketonization. Differently, the longer Ti_5c–Ti_5c distance (3.86 Å) on the (101) and (001) facets reduces the strong lateral repulsion between methyl groups. Notably, C–C coupling on the (101) facet is likely more difficult than on the (001) facet because of the long Ti_5c–Ti_5c distance (5.45 Å) between the two Ti_5c–O_3c–Ti_5c–O_3c rows of (101) facet as well as the steric hindrance resulting from the first layer of O atoms. In contrast, the almost flat surface and square arrangement of the Ti_5c centers of the (001) facet impose little steric hindrance, therefore making it more susceptible to C–C coupling.

Fig. 6 (a) Conversion of propionic acid and selectivity of 3-pentanone on the TiO₂ catalysts as a function of W/F. Reaction conditions: P_total = 101.325 kPa, P_acid = 4.0 kPa, and T = 350 °C.³¹ (b) Integral IR band intensities of monodentate propionate and bidentate propionate on TiO₂(101), TiO₂(100) and TiO₂(001), and the monodentate/bidentate propionate ratio as a function of desorption.³¹ (c) DFT-determined monodentate and bidentate propionates at full coverage on TiO₂(101), (100), and (001) surfaces (PBE + D3). The side and top views are shown in the top and bottom rows, respectively. Light blue, red, gray, and white spheres are Ti, O, C, and H atoms, respectively. The distances in the figure are given in Å. Ti and O atoms and Ti–O bonds in the lower layers are simplified as lines.³¹

CeO₂ is another attractive catalyst for the ketonization reaction.^78–80 Through the synthesis of CeO₂ with different nano-shapes, Guo et al. investigated the effect of CeO₂ facets on the ketonization of propionic acid.⁴¹ The conversion of propionic acid at a space time of 0.05 h on the CeO₂ nano rods (CeO₂-R mainly exposing (110) facet) is much higher than that on nano octahedrons (CeO₂-O mainly exposing (111) facet), and nano cubes (CeO₂-C mainly exposing (100) facet) at 250–400 °C (Fig. 7a). Specifically, the propionic acid conversions on CeO₂-R, CeO₂-O, and CeO₂-C at 350 °C are 65.6, 30.7, and 5.6%, respectively. The measured activation energies (E_a) at 330–370 °C are 124.8, 131.9, and 145.2 kJ mol⁻¹ on CeO₂-R, CeO₂-O, and CeO₂-C, respectively (Fig. 7b), which align well with the catalytic activities. Significantly different characteristics of adsorbed intermediates were observed on different CeO₂ nano shapes in the IR experiments of propionic acid adsorption and desorption. Although monodentate propionate is a minor surface species relative to the bidentate one, the integral band intensity of monodentate propionate on CeO₂-R is 4 times higher than that on CeO₂-O and even 10 times higher than that on CeO₂-C. This trend agrees well with the activity and therefore implies that the monodentate propionate is the active intermediate toward ketonization. The DFT results of the adsorption of propionic acid on various CeO₂ facets (Fig. 8a and b) indicated that the surface coverage of various species is strongly dependent on the fine surface structure of different facets. It is evident that the adsorption energies of both monodentate and bidentate propionates on the (110) facet were not affected by the surface coverages. This can be attributed to O and Ce atoms being in the same layer and Ce atoms loosely arranged on the (110) facet, leading to a slight lateral repulsion between propionates at high coverages. In contrast, although the propionates were slightly stabilized at an intermediate monolayer (ML) coverage of 1/2 on the (111) facet, the monodentate propionate became significantly unstable at high ML coverage of 1. This is due to the tilted surface of the Ce–O pairs, which makes the monodentate propionate occupy more space on the surface. However, the high density of surface Ce cations provides reduced room to accommodate propionate, resulting in strong lateral repulsion between propionates at the high coverages, destabilizing the monodentate propionates. As a result, monodentate propionate on the (110) facet is more favorable than on the (111) facet at high coverages, resulting in higher activity toward ketonization. The unique surface configuration of the (100) facet further disfavors the monodentate adsorption, hence, the lowest ketonization activity. Moreover, the outmost layer O atoms on both (111) and (100) facets of CeO₂ may impose steric hindrance for the formation of the C–C coupling transition state, resulting in lower ketonization activity with respect to the (110) facet, which has both O and Ce atoms in the outmost layer.

Fig. 7 (a) Conversion of propionic acid on the CeO₂ catalyst as a function of temperature. Reaction conditions: P_total = 101.325 kPa, P_acid = 4.0 kPa, W/F = 0.05 h, and the time on stream is 30 min for each temperature.⁴¹ (b) Arrhenius plots of ketonization of propionic acid on CeO₂ catalysts. Reaction conditions: T = 330–370 °C, P_total = 101.325 kPa, P_acid = 4.0 kPa, and the conversion is <10% by adjusting the space time.⁴¹

Fig. 8 (a) DFT-determined monodentate and bidentate propionate on CeO₂(110), (111), and (100) surfaces (PBE + D3). The side and top views are shown in the top and bottom rows, respectively. Light blue, red, gray, and white spheres are Ce, O, C, and H atoms, respectively. The distances in the figure are given in Å. Ce and O atoms and Ce–O bonds in the lower layers are simplified as lines. The acid coverage (θ_acid) is defined as the ratio of adsorbed propionate (n_acid) to exposed Ce–O sites (n_Ce–O): θ_acid = n_acid/n_Ce–O (ML). The full coverage of monodentate and bidentate propionate is 1 and 1/2 ML, respectively, on both (110) and (111) surfaces. The full coverage of monodentate and bidentate propionate is 1/4 and 1/8 ML on the (100) surface, respectively.⁴¹ (b) Adsorption energy of monodentate propionate and bidentate propionate at different coverages on CeO₂(110) and (111) surfaces.⁴¹

Note that the ketonization activity on neither TiO₂ nor CeO₂ with different exposed facets can be simply correlated with the acid–base properties.^31,41 This is likely due to the acid and base sites influencing different elementary steps of ketonization and their effects may cancel each other. Consequently, the fine surface configuration structure of various facets becomes the dominant factor, which influences the adsorption structure of the reactant, the coverage of the intermediate, as well as the reactivity of the intermediate toward ketonization. In particular, one may expect that the fine configuration factors, such as the metal cation distances and the location of surface metal and O atoms in the same or different layers, may influence the ease of formation of the C–C coupling transition state from two adsorbed intermediates on adjacent surface metal cations. Thus, the fine surface configuration structure may play a crucial role in the bimolecular C–C coupling reaction.

3.3. Aldol condensation of ketones and aldehydes

Aldol condensation of ketones is an important organic synthesis reaction that can be applied to upgrade biomass-derived feedstocks.^81,82 This reaction converts two ketones into a longer chain ketone with the formation of a C=C bond, which removes 1/2 O content without consuming H₂. This reaction initiates with the coordination of carbonyl O to the surface coordinatively unsaturated metal cation (Fig. 9). Then, abstraction of α-H of the ketone at the adjacent surface O anion (base site) forms an enolate, which nucleophilically attacks carbonyl C of adjacent ketone to form a new C–C bond. The subsequent dehydration results in the formation of a new ketone with a C=C bond at the β position. Both the enolization and C–C coupling have been proposed to be the rate-limiting step for the aldol condensation of ketones.^83,84

Fig. 9 Proposed elementary steps of acetone aldol condensation on TiO₂.⁸⁵

Lin et al. investigated the conversion of acetone on TiO₂ catalysts at 523 K⁸⁵ and found that the aldol condensation rate on TiO₂ with the dominant (001) facet is ∼1.5 times higher than that on TiO₂ with the dominant (101) facet (Fig. 10a). Also, the E_a derived from Arrhenius plots on TiO₂(001) is lower than that on TiO₂(101) (69 ± 4 and 109 ± 1 kJ mol⁻¹). Kinetic isotope effect experiments indicated that the C–C coupling step is the rate-limiting step. The authors built H₂O-covered TiO₂ facet models to simulate the aldol condensation process via DFT. Under such circumstances, the C–C bond formation takes place between an enolate on the surface Ti cation and another acetone bonded at the surface hydroxyl group on TiO₂ (Fig. 10b). The potential energy profiles illustrated that TiO₂(001) has a much lower barrier of C–C coupling than TiO₂(101) (50 vs. 120 kJ mol⁻¹). The authors attributed this difference to both acid–base properties and geometric structure. On the one hand, the C–C coupling step requires proton transfer from the surface hydroxyl group to the aldol precursor, followed by partial desorption to form the adsorbed aldol. The weaker Lewis acid and Brønsted base strengths of the (001) facet facilitate proton transfer and desorption, thereby enhancing the C–C coupling step on the (001) facet. On the other hand, compared with the (101) facet, the (001) facet has a smoother surface configuration and thus imposes little steric hindrance for intermolecular C–C coupling, which leads to a lower barrier on the (001) facet.

Fig. 10 (a) Rates for acetone aldol condensation on TiO₂(101) + Cu/SiO₂(1 : 2 mass) and TiO₂(001) + Cu/SiO₂(1 : 1 mass) at 523 K as a function of acetone partial pressure (P_one) under constant H₂ partial pressures of 8, 17, 34, and 68 kPa. (b) Potential energy profiles and intermediate structures of acetone condensation on (001) (orange) and (101) (green) facets of anatase TiO₂, with zero energy corresponding to two acetone molecules in the gas phase away from the TiO₂ surface.⁸⁵

CeO₂ with varying facets has also been investigated for aldol condensation of ketone. Li et al. examined the aldol condensation of cyclopentanone to the dimeric product 2-cyclopentylidenecyclopentanone on CeO₂ nano-shapes at 200 °C and 200 psi.⁸⁶ The condensation rate with a cyclopentanone concentration of 1.5 mol L⁻¹ for CeO₂-R was 168.5 mmol g_cata⁻¹ h⁻¹, whereas they were only 10.9 and 23.9 mmol g_cata⁻¹ h⁻¹ for CeO₂-C and CeO₂-O, respectively (Fig. 11a). Moreover, kinetic fittings demonstrated that the reaction rate on CeO₂-R can be fitted with the first-order Langmuir–Hinshelwood (L–H) model, whereas CeO₂-C and CeO₂-O match the second-order L–H model (Fig. 11b). The authors suggested that such different kinetic behaviors are related to the surface fine configuration structure. That is, the first layer of O atoms on the (100) and (111) facets may impose steric hindrance for intermolecular C–C coupling. Consequently, the overall aldol condensation rates on CeO₂-C and CeO₂-O are controlled by the C–C coupling step, which follows the second-order reaction kinetics. In contrast, the outmost layer contains both Ce and O atoms in the (110) facet of CeO₂-R, which imposes little steric hindrance for the intermolecular C–C coupling step. Thus, the C–C coupling is unlikely to be the rate-limiting step. Instead, the reaction may be limited by the α-H abstraction step and therefore shows a first-order kinetic behavior.

Fig. 11 (a) Cyclopentanone condensation rates on CeO₂ nanoshapes with respect to the initial concentration of cyclopentanone at 200 °C and 200 psi. (b) Reaction rate data at 200 °C and 200 psi fitting from the first-order Langmuir–Hinshelwood model, second-order Langmuir–Hinshelwood model and second-order Eley–Rideal model, respectively.⁸⁶

Similar to the conversion of ketones, the aldol condensation reaction can also be used to convert aldehydes, producing a longer chain α,β-unsaturated aldehyde. As the reaction mechanisms for the aldol condensation of ketones and aldehydes are the same, metal oxides can be applied for the conversion of aldehydes. The effect of crystal facets on the aldol condensation of aldehydes has also been studied. Mann et al. studied the adsorption and coupling of acetaldehyde to crotonaldehyde over different CeO₂ nanoshapes (wires (110), cubes (100) and octahedrons (111)).⁸⁷ The conversion of acetaldehyde decreased following the order of wires > cubes > octahedrons at both 300 and 400 °C (Fig. 12a and b). Specifically, the initial acetaldehyde conversion was >90% on wires at 400 °C, which was much higher than that of cubes (∼60%) and octahedrons (∼40%). The authors suggested that the lower coordination number of surface O atoms on (100) in comparison to (111), i.e., the O_2c and O_3c on (100) and (111), respectively, favor α-H abstraction (enolization) and thus facilitates the aldol condensation reaction. In addition, the authors also implied that the highest density of oxygen vacancy in wires, resulting from the lowest oxygen vacancy formation energy on the (110) facet, led to more active sites for enolization,^88,89 and therefore, the highest activity on the (110) facet. It should be noted that the higher activity on (110) and (100) facets also led to higher selectivity toward side products, particularly at a higher reaction temperature of 400 °C, while the (111) facet showed higher stability and higher selectivity of crotonaldehyde. Despite the authors ascribing the activity difference to the acid–base property and density of oxygen vacancy, one may anticipate that the surface configuration structure also affects the activity. That is, the almost flat surface with both Ce and O in the outmost layer of (110) imposes little steric hindrance on the intermolecular C–C coupling, resulting in the highest activity.

Fig. 12 Conversion of acetaldehyde in a gas stream of 0.5% acetaldehyde/He on CeO₂ cubes, wires, and octahedrons with respect to the reaction time at (a) 400 °C and (b) 300 °C.⁸⁷

3.4. Other facet-dependent reactions beyond the conversion of oxygenates

Although this perspective focuses on the conversion of biomass-derived oxygenates, crystal-facet engineering of metal oxide catalysts can be applied to many other reactions. For example, the oxidation of carbon monoxide,⁹⁰ hydrogen chloride³⁸ and selective oxidation of organic compounds^60,91–93 have also been reported to be affected by various morphologies of CeO₂ (nano rod, nano cube, nano octahedron, nano particle and nano polyhedron), and can be well correlated with the highly unsaturated surface coordination and redox property of various facets. This is because the oxidation reactions involve O₂ adsorption and activation at the oxygen vacancy sites as well as transfer of the surface reactive O to reactant to accomplish the oxidation.^94,95 The hydrogenation of acetylene⁹⁶ and propylene⁹⁷ was also found to be dependent on the morphology of CeO₂, likely related to the varied H₂ heterolytic dissociation and reactant adsorption properties on various facets.

3.5. Brief discussion on the structure–activity relationship

The above sections surveyed the conversion of biomass-derived oxygenates, i.e., alcohols, carboxylic acids, ketones, and aldehydes, using nano crystal metal oxides with well-defined facets. Despite a certain reaction that may be more closely correlated with a certain facet-dependent property caused by the effect of morphology, some general rules of structure–activity relationship via crystal facet engineering can be reached. (1) The acid–base property is the primary facet-dependent property, which influences all elementary steps of metal oxide-mediated reaction through interaction with reactants and intermediates. A clear correlation between alcohol dehydration/dehydrogenation and the acid–base property was demonstrated^70,71 since this reaction is a simple unimolecular reaction and is only dependent on the acid–base property. However, no such clear correlation is found for some more complex bimolecular reactions, such as ketonization, which likely resulted from the effects of the acid–base property in different elementary steps that cancelled each other. (2) The fine surface geometrical structure influences the distribution of surface metal cations and O anions and was found to have a good correlation with the bimolecular reactions of ketonization and aldol condensation. On the one hand, the fine surface structure influences the fine adsorption configuration structures of reactants and intermediates, as well as their coverages. On the other hand, the fine surface structure may impose steric hindrance for intermolecular C–C coupling, which thus affects the barrier for C–C coupling. In general, the flatter facets, such as CeO₂(110) and TiO₂(001), show higher activity than sawtooth facets, such as CeO₂(111) and TiO₂(101), for these bimolecular reactions. (3) The presence of oxygen vacancy lowers the coordination number of vicinal metal cations, which alters the acid–base properties. Consequently, the facet-dependent density of oxygen vacancies alters the adsorption and activation of reactants and eventually affects the activity. The presence of oxygen vacancy favors the aldehyde adsorption in the η-2 configuration, compared to the η-1 configuration on the facet without the vacancy.^87,89 (4) The presence of dissociated or molecular H₂O alters the density and strength of the acid–base property and hence also influences the activity. Clearly, the surface H₂O is dependent on both facet and reaction conditions. We note that these factors are somehow interconnected and may synergistically influence a certain reaction. In some cases, decoupling these factors is difficult.

4. Crystal-facet engineering of metal oxides

Synthesis of metal oxide nanocrystals with different morphologies (and therefore different fractions of exposed facets) has become a booming area of nanomaterials. In addition to the nanocrystals discussed in section 2, various kinds of nanocrystals with different morphology, size, and surface properties have been synthesized and explored in various reactions. In this section, we briefly summarize the advances in crystal-facet engineering of TiO₂ and CeO₂ nanocrystals, as well as the application in other reactions beyond those discussed in section 3.

4.1. Controlling the morphology and size of nanocrystals

The morphology of nanocrystals directly affects the exposing facets and their contents. Table 1 summarizes the recently reported nanocrystals with different morphologies as well as their mainly exposed facets. TiO₂ nano belts with the dominant (101) facet were synthesized by calcining titanate at 700 °C, which showed excellent photocatalytic activity for the generation of the superoxide radical (O₂⁻) from O₂.⁹⁸ Based on the agglomeration growth process of nano bipyramids, TiO₂ microspheres consisting of major (001) and minor (101) facets were synthesized by the hydrothermal hydrolysis of TiB₂.⁹⁹ In addition, TiO₂ nanostructures exposing high-index facets, such as micro polyhedrons exposing mainly the (105) facet, and nano needles exposing (201) and (401) facets, have also been reported.^100,101 The synthesis of one-dimensional (1D) nanostructures (wires and tubes) is a popular area in CeO₂-based nanomaterials. Tang et al. reported a surfactant-free route to synthesize CeO₂ nano wires (diameters 20–70 nm and lengths up to 40 μm), mainly exposing the (200) facet.¹⁰² Han et al. developed a hydrothermal route to produce CeO₂ nano tubes, mainly exposing the (111) facet at low temperature and pressure.¹⁰³ In addition, the synthesis of CeO₂ nano tubes by a simple oxidation–coordination-assisted dissolution process of the Ce(OH)₃ nano rods was also reported, which produced CeO₂ nano tubes (mainly exposing the (111) facet) with large cavities and thin walls.¹⁰⁴

Table 1 Morphologies and structural parameters and exposed facets of TiO₂ and CeO₂

Metal oxides	Morphologies	Structural parameters	Exposed crystal facets	Ref.
Anatase TiO2	Nano bipyramid	∼15 nm in length, 12 nm in width	97.6% (101) and 2.4% (001)	31
		80–125 nm in length, 30–50 nm in width	90% (101) and 10% (001)	70
		∼60 nm in length, 30 nm in width	98% (101) and 2% (001)	85
		30–50 nm in length, 20–30 nm in width	>90% (101) and <10% (001)	9
Anatase TiO2	Nano sheets	∼64 nm in width, 8 nm in thickness	72.6% (001) and 27.4% (101)	31
		150–200 nm in width, 40–60 nm in thickness	76% (001) and 24% (101)	70
		∼400 nm in width, 150 nm in thickness	65% (001) and 35% (101)	85
		∼250 nm in width, 125 nm in thickness	46% (001) and 54% (101)	9
Anatase TiO2	Nano rods	∼28 nm in length, 9 nm in width	77.8% (100), 19.3% (101) and 10% (001)	31
		70 nm in length, 15 nm in width	78.8% (100), 20.2% (101) and 1.0% (001)	30
		∼29 nm in length, 12 nm in width	72% (100), others are (101) and (001)	105
Anatase TiO2	Nano belts	60–400 nm in width, 0–30 μm in length and ∼10 nm in thickness	Major (101) and minor (010)	98
Anatase TiO2	Micro spheres	∼1.5 μm in radius	Major (001) and minor (101)	99
Anatase TiO2	Micro polyhedrons	∼2.4 μm in length, ∼1.2 μm in thickness	Major (105) and minor (101)	100
Anatase TiO2	Nano needles	∼350 nm in length, 30 nm in width	67% (401) and 33% (201)	101
Cubic CeO2	Nano octahedrons	∼15 nm in length and width	(111)	41
Cubic CeO2	Nano rods	∼100 nm in length, 11 nm in width	(100) and (110)	41
		∼100 nm in length, 5–10 nm in width	(100), (110) and (111)	106
		∼200 nm in length, 5–10 nm in width	Major (111) and minor (110)	107
		100–300 nm in length, 12–20 nm in width	Major (111), (100) and minor (110)	94
Cubic CeO2	Nano cubes	∼20 nm in length and width	Major (100) and minor (111)	41
Cubic CeO2	Nano wires	Several μm in length, ∼15 nm in diameter	(200)	102
Cubic CeO2	Nano tubes	Several μm in length, 5–30 nm in diameter	(111)	103
		∼100 nm in length, 15–25 nm in diameter		104

---

Notably, besides the morphology, the crystal size of metal oxide nanocrystals is also a key factor that determines the surface property of metal oxides. It is apparent that altering the size of a certain nanocrystal will affect the structural parameters, such as the fraction of exposed facets. Take TiO₂ nano sheets as an example (Table 1), it is clear that the fraction of exposed crystal facets varied as a function of crystal size reported in different works. Moreover, it has been demonstrated that the concentration of oxygen vacancies varied when the crystal size was changed.^108–110 Hence, developing a method to precisely synthesize nanocrystals with the target morphology and size is crucial for crystal-facet engineering as well as application in various reactions.

4.2. Surface modification of nanocrystals

To enhance the performance of metal oxide nanocrystals with varying exposed facets, surface modification of nanocrystals has been widely used. A number of approaches to modify the crystal facets, including doping other metal oxides, loading other metals or metal oxides, and varying pretreatment, have been proposed.

Doping an appropriate amount of metal (oxide) into the lattice of CeO₂ or TiO₂ can form a solid solution without altering the crystal structures and morphologies. This approach has been proven to be an effective method for surface modification. Even a small amount of metal doping can alter the electronic structure and surface properties of the crystal facets. The disparity between the ionic radius of doped metal ions and host ions (Ce⁴⁺ or Ti⁴⁺) results in the formation of Ce(Ti)–O–M local structures within the crystal lattice.¹¹¹ These local structures induce lattice strain, which favors the formation of more oxygen vacancies, enhances ionic mobility and influences catalytic activity, selectivity, and stability. Liu et al. synthesized Zr-doped CeO₂ nano rods with different Zr/Ce ratios for dimethyl carbonate (DMC)synthesis from CO₂ and methanol (CO₂ + 2CH₃OH → CH₃OCOOCH₃ + H₂O).¹¹² The authors suggested that the formation of DMC occurs through coupling two CH₃OH (breaking the C–O and O–H bonds, respectively) with a CO₂ adsorbed on the oxygen vacancy, and the activity is influenced by the abundance of surface oxygen vacancies. Compared with CeO₂ nano rods, the activity of Zr-doped CeO₂ nano rods with the highest amount of oxygen vacancies increased by ∼50%.

Loading metals or metal oxides on TiO₂ or CeO₂ with predominately exposed facets is also an attractive approach to surface modification, which creates an interface and tunes metal–support interactions.^{105,113–115} This kind of surface modification is widely explored for a variety of reactions. For example, Wu et al. loaded Pt on TiO₂ with different dominant exposed facets and found that the degree of strong metal–support interaction (SMSI) decreases following the order of Pt/TiO₂(100) > Pt/TiO₂(001) > Pt/TiO₂(101).¹⁰⁵ The maximum density of interfacial Pt–Ov–Ti³⁺ sites was achieved on TiO₂(001) with medium SMSI, resulting in the highest reaction rate for hydrodeoxygenation of m-cresol to toluene. Si et al. loaded Au on CeO₂ rods, cubes and polyhedrals, which were used for investigating the water gas shift (WGS) reaction (CO + H₂O ↔ CO₂ + H₂).¹¹⁵ They demonstrated that the CeO₂ rods with mainly exposed (110) have the highest density of oxygen vacancies, which facilitate anchoring and dispersing Au clusters, resulting in the highest WGS activity.

In addition, modification of metal oxide nanostructures via different treatments, such as chemical modification, plasma treatment modification and irradiation modification,¹¹⁶ to create defect sites is also possible. For instance, Gao et al. developed a redox chemical etching method to regulate the surface properties of CeO₂ nano rods (with mainly CeO₂(110) facets).¹¹⁷ The etched nano rods enhanced the specific surface area, density of oxygen vacancies and surface Ce³⁺, which promote O₂ adsorption and activation at the oxygen vacancies and thus improve the performance of CO oxidation. Yang et al. performed plasma treatment of CeO₂ nano flakes (mainly (111) facet) using Ar, H₂, O₂, and NH₃ and showed that the enhancement in density of surface Ce³⁺ and oxygen vacancies is more favorable when the reducing plasma atmosphere is applied.¹¹⁸ E. Krumov et al. employed excimer laser processing to modify the surface of CeO₂(111) thin films, leading to an oxygen-deficient CeO₂ surface via reduction.¹¹⁹

5. Summary and outlook

In this perspective, we have surveyed the surface fine structure and properties of various facets of TiO₂ and CeO₂, as well as the conversion of biomass-derived oxygenates (alcohol, carboxylic acid, ketone, and aldehyde) on these facets of nano crystal catalysts. Significant progress in understanding the structure–activity relationship has been achieved from experimental and theoretical studies via crystal-facet engineering of metal oxides. On the one hand, relationships between structure and activity can be elucidated via facet-dependent acid–base properties, geometric structure, and the presence of oxygen vacancies and H₂O. On the other hand, the complex interplay between these factors makes the elaboration of structure–activity relationships elusive.

Crystal-facet engineering represents an attractive approach for integrating experimental and theoretical works, offering a reliable basis for refining theoretical modelling and identifying reaction pathways. It can further be combined with surface engineering, such as loading noble metals, doping second metals and introducing defects, to design optimal catalysts. Hence, crystal-facet engineering represents a frontier in reaction chemistry and catalysis engineering. Despite extensive scientific interest as well as significant advantages of crystal-facet engineering, the synthesis of nano crystals with predominantly exposed facets and a comprehensive understanding of the facet of metal oxide-mediated reactions from both experimental and theoretical aspects remain limiting. Significant obstacles still exist in developing advanced catalysts via crystal-facet engineering. These challenges reflect the complexities and difficulties involved in optimizing reaction performance through crystal-facet engineering. In the future, the following aspects deserve further investigation.

(1) Development of novel synthesis methods to precisely control the percentage of a specific facet and the crystal size of nanocrystal catalyst. Although the synthesis of nano crystal catalyst with predominantly exposed facet is not the focus of this perspective, precisely controlling the predominantly exposed facet, particularly for those with high surface energies, is a crucial aspect in crystal-facet engineering. For example, during the synthesis of TiO₂ nano crystals with a predominant (001) facet, the (101) facet was also introduced simultaneously.^9,70 Thus, developing simple preparation methods to improve the fraction of the target facet is highly desirable to minimize the contribution of other facets on the catalytic performance and can provide a more clear and accurate facet dependent structure–activity relationship. Moreover, the advanced method that can precisely control the crystal size of nanocrystals is also highly desirable.

(2) Application of in situ techniques to monitor crystal facets under reaction conditions. The possible surface reconstruction under real conditions may alter the surface fraction of exposing facets and thus may render an unreliable correlation between the facet and activity performance. Surface reconstruction is currently largely ignored in crystal facet engineering studies. Thus, it is necessary to in situ monitor the facets of nano crystal catalysts under real reaction conditions and to exclude the contribution from reconstructed facets. To this end, the application of in situ technology is essential. The in situ infrared technique coupled with a probe molecule is well established. Meanwhile, it is anticipated that near-ambient-pressure X-ray photoelectron spectroscopy (XPS), near-ambient-pressure scanning or transmission electron microscopy (SEM/TEM) and in situ low-energy ion scattering (LEIS) may play an important role in the future research.

(3) Application of advanced theoretical simulation methods. First-principles DFT calculations have become fundamental for studying catalytic processes on various surfaces. Under real reaction conditions, the surface composition and structure of catalysts are highly complex, such as the presence of defects (oxygen vacancies), surface H₂O and hydroxyl groups, and dynamic changes of the active site. To address these challenges, the application of more advanced theoretical methods is essential. The dynamics simulations and machine learning are promising avenues for advancing the understanding of metal oxide facet-dependent reactions.

(4) Understanding the metal oxide facet-dependent structure–activity relationship. Crystal-facet engineering provides a powerful approach to exploring the structure–activity relationships of reactions mediated by metal oxide catalysts. However, the accurate understanding of the facet-dependent structure–activity relationship is limited by the heterogeneity of the synthesized nano crystal catalyst and the real surface structure of the facet under reaction conditions, i.e., the presence of oxygen vacancies and surface H₂O and the reconstruction of facets. Therefore, it is imperative to investigate the structure–activity relationships via crystal-facet engineering through the implementation of advanced synthesis methods, in situ characterization techniques, as well as advanced theoretical calculation methods. This is of great importance for the rational design of metal oxide catalysts for various reactions via crystal-facet engineering.

(5) Exploration of other facet-dependent descriptors to correlate with catalytic reaction performance. In this perspective, we mainly discussed the facet-dependent acid–base property and geometrical configuration, as well as its relationship with catalytic performance. These facet-dependent factors can be determined from both experimental or computational techniques. Some other facet-dependent descriptors, such as density of oxygen vacancies, density of surface H₂O or –OH groups, electronic structure, and coordination number, may be further explored and correlated with various reactions mediated by metal oxides, which may facilitate the rational design of metal oxide catalysts for various reactions beyond the conversion of oxygenates.

In conclusion, crystal-facet engineering of metal oxides offers promising avenues for reaction chemistry and catalytic reaction engineering mediated by metal oxides, which is particularly suitable for the conversion of biomass-derived oxygenates. However, significant challenges remain and deserve further investigation on the rational design of metal oxide catalysts and the mediated reactions and processes.

Data availability

No new data are present in this perspective.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors acknowledge the financial support from the National Natural Science Foundation of China (22278299) and the Fundamental Research Funds for the Central Universities.

References

1. J. M. Vohs, Chem. Rev., 2012, 113, 4136–4163.
2. J. T. Grant, J. M. Venegas, W. P. McDermott and I. Hermans, Chem. Rev., 2017, 118, 2769–2815.
3. W. Zhou, Y. Tang, X. Zhang, S. Zhang, H. Xue and H. Pang, Coord. Chem. Rev., 2023, 477, 214949–214970.
4. J. C. Védrine, Chin. J. Catal., 2019, 40, 1627–1636.
5. Y. Li, D. He, Q. Zhu, X. Zhang and B. Xu, J. Catal., 2004, 221, 584–593.
6. S. Ding, H. Wang, J. Han, X. Zhu and Q. Ge, Ind. Eng. Chem. Res., 2018, 57, 17086–17096.
7. E. Iglesia, D. G. Barton, J. A. Biscardi, M. J. L. Gines and S. L. Soled, Catal. Today, 1997, 38, 339–360.
8. Z. Liu, Y. Guo, H. Wang and X. Zhu, Appl. Catal., A, 2024, 672, 119593.
9. Y. Wu, F. Gao, H. Wang, L. Kovarik, B. Sudduth and Y. Wang, J. Phys. Chem. C, 2021, 125, 3988–4000.
10. S. Wang and E. Iglesia, J. Am. Chem. Soc., 2018, 140, 775–782.
11. C. Zhao, C. Watt, P. R. Kent, S. H. Overbury, D. R. Mullins, F. C. Calaza, A. Savara and Y. Xu, J. Phys. Chem. C, 2018, 123, 8273–8286.
12. D. R. Mullins, P. M. Albrecht, T.-L. Chen, F. C. Calaza, M. D. Biegalski, H. M. Christen and S. H. Overbury, J. Phys. Chem. C, 2012, 116, 19419–19428.
13. P. M. Jayaweera, E. L. Quah and H. Idriss, J. Phys. Chem. C, 2007, 111, 1764–1769.
14. D. R. Mullins, P. M. Albrecht and F. Calaza, Top. Catal., 2013, 56, 1345–1362.
15. P. Thomas, C. W. Lai and M. R. Bin Johan, J. Anal. Appl. Pyrolysis, 2019, 140, 54–85.
16. B. Boekaerts and B. F. Sels, Appl. Catal., B, 2021, 283, 119607–119643.
17. J. Asomaning, S. Haupt, M. Chae and D. C. Bressler, Renewable Sustainable Energy Rev., 2018, 92, 642–657.
18. N. A. Huq, G. R. Hafenstine, X. Huo, H. Nguyen, S. M. Tifft, D. R. Conklin, D. Stück, J. Stunkel, Z. Yang, J. S. Heyne, M. R. Wiatrowski, Y. Zhang, L. Tao, J. Zhu, C. S. McEnally, E. D. Christensen, C. Hays, K. M. Van Allsburg, K. A. Unocic, H. M. Meyer, Z. Abdullah and D. R. Vardon, Proc. Natl. Acad. Sci. U. S. A., 2021, 118, e2023008118.
19. M. I. Zaki, M. A. Hasan and L. Pasupulety, Langmuir, 2001, 17, 768–774.
20. M. A. Hasan, M. I. Zaki and L. Pasupulety, J. Phys. Chem. B, 2002, 106, 12747–12756.
21. M. A. Hasan, M. I. Zaki and L. Pasupulety, Appl. Catal., A, 2003, 243, 81–92.
22. X. Zhao, M. D. Susman, J. D. Rimer and P. Bollini, ChemCatChem, 2020, 13, 6–27.
23. R. Katal, S. Masudy-Panah, M. Tanhaei, M. H. D. A. Farahani and H. Jiangyong, Chem. Eng. J., 2020, 384, 123384–123409.
24. Y. Zhu, C. Chen, P. Cheng, J. Ma, W. Yang, W. Yang, Y. Peng, Y. Huang, S. Zhang and G. Seong, Dalton Trans., 2022, 51, 6506–6518.
25. W. Chen, Q. Kuang, Q. Wang and Z. Xie, RSC Adv., 2015, 5, 20396–20409.
26. U. Diebold, Surf. Sci. Rep., 2003, 48, 53–229.
27. G. Liu, H. G. Yang, J. Pan, Y. Q. Yang, G. Q. Lu and H.-M. Cheng, Chem. Rev., 2014, 114, 9559–9612.
28. S. Chen, D. Li, Y. Liu and W. Huang, J. Catal., 2016, 341, 126–135.
29. X. G. Han, Q. Kuang, M. S. Jin, Z. X. Xie and L. S. Zheng, J. Am. Chem. Soc., 2009, 131, 3152–3153.
30. L. Liu, X. Gu, Y. Cao, X. Yao, L. Zhang, C. Tang, F. Gao and L. Dong, ACS Catal., 2013, 3, 2768–2775.
31. J. Huang, L. Li, X. Wu, Y. Guo, Z. Yang, H. Wang, Q. Ge and X. Zhu, Catal. Sci. Technol., 2023, 13, 5924–5937.
32. G. Li, K. Fang, Y. Ou, W. Yuan, H. Yang, Z. Zhang and Y. Wang, Prog. Nat. Sci.: Mater. Int., 2021, 31, 1–13.
33. Y. A. Çetin, B. Martorell, F. Serratosa, N. Aguilera-Porta and M. Calatayud, J. Phys.: Condens. Matter, 2022, 34, 314004–314013.
34. W. T. Yuan, B. E. Zhu, X. Y. Li, T. W. Hansen, Y. Ou, K. Fang, H. S. Yang, Z. Zhang, J. B. Wagner, Y. Gao and Y. Wang, Science, 2020, 367, 428–430.
35. R. Erdogan, O. Ozbek and I. Onal, Surf. Sci., 2010, 604, 1029–1033.
36. M. Sumita, C. P. Hu and Y. Tateyama, J. Phys. Chem. C, 2010, 114, 18529–18537.
37. J. Blomquist, L. E. Walle, P. Uvdal, A. Borg and A. Sandell, J. Phys. Chem. C, 2008, 112, 16616–16621.
38. C. Li, Y. Sun, I. Djerdj, P. Voepel, C.-C. Sack, T. Weller, R. Ellinghaus, J. Sann, Y. Guo, B. M. Smarsly and H. Over, ACS Catal., 2017, 7, 6453–6463.
39. H. X. Mai, L. D. Sun, Y. W. Zhang, R. Si, W. Feng, H. P. Zhang, H. C. Liu and C. H. Yan, J. Phys. Chem. B, 2005, 109, 24380–24385.
40. P. Sudarsanam, B. Hillary, M. H. Amin, S. B. A. Hamid and S. K. Bhargava, Appl. Catal., B, 2016, 185, 213–224.
41. Y. Guo, Y. Qin, H. Liu, H. Wang, J. Han, X. Zhu and Q. Ge, ACS Catal., 2022, 12, 2998–3012.
42. M. Molinari, S. C. Parker, D. C. Sayle and M. S. Islam, J. Phys. Chem. C, 2012, 116, 7073–7082.
43. Z. Ren, N. Liu, B. Chen, J. Li and D. Mei, J. Phys. Chem. C, 2018, 122, 4828–4840.
44. M. Nolan, S. C. Parker and G. W. Watson, Surf. Sci., 2005, 595, 223–232.
45. H. Metiu, S. Chrétien, Z. Hu, B. Li and X. Sun, J. Phys. Chem. C, 2012, 116, 10439–10450.
46. Y. Wang, F. Wang, Q. Song, Q. Xin, S. Xu and J. Xu, J. Am. Chem. Soc., 2013, 135, 1506–1515.
47. S. Zhang, Z.-Q. Huang, Y. Ma, W. Gao, J. Li, F. Cao, L. Li, C.-R. Chang and Y. Qu, Nat. Commun., 2017, 8, 15266–15276.
48. W. Yuan, Y. Wang, H. Li, H. Wu, Z. Zhang, A. Selloni and C. Sun, Nano Lett., 2015, 16, 132–137.
49. Y. Wang, H. Sun, S. Tan, H. Feng, Z. Cheng, J. Zhao, A. Zhao, B. Wang, Y. Luo, J. Yang and J. G. Hou, Nat. Commun., 2013, 4, 2214–2221.
50. R. Hengerer, B. Bolliger, M. Erbudak and M. Grätzel, Surf. Sci., 2000, 460, 162–169.
51. M. Lazzeri and A. Selloni, Phys. Rev. Lett., 2001, 87, 266105–266108.
52. S. Selçuk and A. Selloni, J. Phys. Chem. C, 2013, 117, 6358–6362.
53. H. Nörenberg and G. A. D. Briggs, Surf. Sci., 1999, 433–435, 127–130.
54. R. E. Tanner, M. R. Castell and G. A. D. Briggs, Surf. Sci., 1998, 412–413, 672–681.
55. P. J. Hardman, N. S. Prakash, C. A. Muryn, G. N. Raikar, A. G. Thomas, A. F. Prime, G. Thornton and R. J. Blake, Phys. Rev. B: Condens. Matter Mater. Phys., 1993, 47, 16056–16059.
56. M. Bugnet, S. H. Overbury, Z. L. Wu and T. Epicier, Nano Lett., 2017, 17, 7652–7658.
57. Y. Pan, N. Nilius, C. Stiehler, H. J. Freund, J. Goniakowski and C. Noguera, Adv. Mater. Interfaces, 2014, 1, 1400404–1400409.
58. Y. Lin, Z. Wu, J. Wen, K. R. Poeppelmeier and L. D. Marks, Nano Lett., 2013, 14, 191–196.
59. S. Colussi, A. Gayen, M. Farnesi Camellone, M. Boaro, J. Llorca, S. Fabris and A. Trovarelli, Angew. Chem., Int. Ed., 2009, 48, 8481–8484.
60. M. Li, Z. Wu and S. H. Overbury, J. Catal., 2013, 306, 164–176.
61. Z. Wu, M. Li and S. H. Overbury, J. Catal., 2012, 285, 61–73.
62. J. Paier, C. Penschke and J. Sauer, Chem. Rev., 2013, 113, 3949–3985.
63. W. Alharbi, E. Brown, E. F. Kozhevnikova and I. V. Kozhevnikov, J. Catal., 2014, 319, 174–181.
64. J. Macht, M. J. Janik, M. Neurock and E. Iglesia, J. Am. Chem. Soc., 2008, 130, 10369–10379.
65. M. Zhang and Y. Yu, Ind. Eng. Chem. Res., 2013, 52, 9505–9514.
66. P. Kostestkyy, J. Yu, R. J. Gorte and G. Mpourmpakis, Catal. Sci. Technol., 2014, 4, 3861–3869.
67. P. Kostetskyy and G. Mpourmpakis, Catal. Sci. Technol., 2015, 5, 4547–4555.
68. M. W. Nolte and B. H. Shanks, Energy Technol., 2016, 5, 7–18.
69. K. Larmier, C. Chizallet, N. Cadran, S. Maury, J. Abboud, A.-F. Lamic-Humblot, E. Marceau and H. Lauron-Pernot, ACS Catal., 2015, 5, 4423–4437.
70. F. Lin, Y. Chen, L. Zhang, D. Mei, L. Kovarik, B. Sudduth, H. Wang, F. Gao and Y. Wang, ACS Catal., 2020, 10, 4268–4279.
71. B. Sudduth, D. Yun, J. Sun and Y. Wang, J. Catal., 2021, 404, 96–108.
72. S. Wang and E. Iglesia, J. Catal., 2017, 345, 183–206.
73. Z. Yang, Q. Yu, H. Wang, Q. Ge and X. Zhu, J. Catal., 2024, 429, 115247.
74. B. Boekaerts, W. Lorenz, J. Van Aelst and B. F. Sels, Appl. Catal., B, 2022, 305, 121052–121066.
75. T. N. Pham, D. Shi and D. E. Resasco, J. Catal., 2014, 314, 149–158.
76. M. Yazdanpanah, M. Fereidooni, V. Marquez, C. V. Paz, T. Saelee, M. S. Villanueva, M. Rittiruam, P. Khajondetchairit, S. Praserthdam and P. Praserthdam, ChemSusChem, 2024, 17, e202301033.
77. J. Pittayatornkul, T. Maluangnont, S. Jongpatiwut, P. Praserthdam, M. Ogawa and T. Sooknoi, React. Chem. Eng., 2024, 9, 2345–2357.
78. O. Nagashima, S. Sato, R. Takahashi and T. Sodesawa, J. Mol. Catal. A: Chem., 2005, 227, 231–239.
79. R. W. Snell and B. H. Shanks, ACS Catal., 2014, 4, 512–518.
80. Y. Guo, Q. Yu, H. Fang, H. Wang, J. Han, Q. Ge and X. Zhu, Ind. Eng. Chem. Res., 2020, 59, 17269–17278.
81. M. J. Climent, A. Corma and S. Iborra, Green Chem., 2014, 16, 516–547.
82. L. Wu, T. Moteki, A. A. Gokhale, D. W. Flaherty and F. D. Toste, Chem, 2016, 1, 32–58.
83. S. Wang, K. Goulas and E. Iglesia, J. Catal., 2016, 340, 302–320.
84. H. Zhang, M. Y. S. Ibrahim and D. W. Flaherty, J. Catal., 2018, 361, 290–302.
85. F. Lin, H. Wang, Y. Zhao, J. Fu, D. Mei, N. R. Jaegers, F. Gao and Y. Wang, JACS Au, 2020, 1, 41–52.
86. G. Li, B. Wang and D. E. Resasco, ACS Catal., 2020, 10, 5373–5382.
87. A. K. P. Mann, Z. Wu, F. C. Calaza and S. H. Overbury, ACS Catal., 2014, 4, 2437–2448.
88. T. L. Chen and D. R. Mullins, J. Phys. Chem. C, 2011, 115, 3385–3392.
89. F. C. Calaza, Y. Xu, D. R. Mullins and S. H. Overbury, J. Am. Chem. Soc., 2012, 134, 18034–18045.
90. K. Zhou, X. Wang, X. Sun, Q. Peng and Y. Li, J. Catal., 2005, 229, 206–212.
91. M. Tinoco, S. Fernandez-Garcia, A. Villa, J. M. Gonzalez, G. Blanco, A. B. Hungria, L. Jiang, L. Prati, J. J. Calvino and X. Chen, Catal. Sci. Technol., 2019, 9, 2328–2334.
92. K. Deori, C. Kalita and S. Deka, J. Mater. Chem. A, 2015, 3, 6909–6920.
93. Y. Li, Z. Wei, F. Gao, L. Kovarik, R. A. L. Baylon, C. H. F. Peden and Y. Wang, ACS Catal., 2015, 5, 3006–3012.
94. X. W. Liu, K. B. Zhou, L. Wang, B. Y. Wang and Y. D. Li, J. Am. Chem. Soc., 2009, 131, 3140–3141.
95. B. Choudhury and A. Choudhury, Mater. Chem. Phys., 2012, 131, 666–671.
96. G. Vilé, S. Colussi, F. Krumeich, A. Trovarelli and J. Pérez-Ramírez, Angew. Chem., Int. Ed., 2014, 53, 12069–12072.
97. E. W. Zhao, H. Zheng, R. Zhou, H. E. Hagelin-Weaver and C. R. Bowers, Angew. Chem., Int. Ed., 2015, 54, 14270–14275.
98. N. Q. Wu, J. Wang, D. Tafen, H. Wang, J. G. Zheng, J. P. Lewis, X. G. Liu, S. S. Leonard and A. Manivannan, J. Am. Chem. Soc., 2010, 132, 6679–6685.
99. G. Liu, J. Pan, L. Yin, J. T. S. Irvine, F. Li, J. Tan, P. Wormald and H. M. Cheng, Adv. Funct. Mater., 2012, 22, 3233–3238.
100. H. B. Jiang, Q. Cuan, C. Z. Wen, J. Xing, D. Wu, X. Q. Gong, C. Li and H. G. Yang, Angew. Chem., Int. Ed., 2011, 50, 3764–3768.
101. H. B. Wu, J. S. Chen, X. W. Lou and H. H. Hng, Nanoscale, 2011, 3, 4082–4084.
102. B. Tang, L. Zhuo, J. Ge, G. Wang, Z. Shi and J. Niu, Chem. Commun., 2005, 3565–3567.
103. W.-Q. Han, L. Wu and Y. Zhu, J. Am. Chem. Soc., 2005, 127, 12814–12815.
104. K. Zhou, Z. Yang and S. Yang, Chem. Mater., 2007, 19, 1215–1217.
105. X. Wu, C.-j. Liu, H. Wang, Q. Ge and X. Zhu, J. Catal., 2023, 418, 203–215.
106. S. Chen, T. Cao, Y. Gao, D. Li, F. Xiong and W. Huang, J. Phys. Chem. C, 2016, 120, 21472–21485.
107. S. Agarwal, L. Lefferts, B. L. Mojet, D. A. J. M. Ligthart, E. J. M. Hensen, D. R. G. Mitchell, W. J. Erasmus, B. G. Anderson, E. J. Olivier, J. H. Neethling and A. K. Datye, ChemSusChem, 2013, 6, 1898–1906.
108. S. Tsunekawa, R. Sahara, Y. Kawazoe and K. Ishikawa, Appl. Surf. Sci., 1999, 152, 53–56.
109. X. D. Zhou and W. Huebner, Appl. Phys. Lett., 2001, 79, 3512–3514.
110. R. K. Hailstone, A. G. DiFrancesco, J. G. Leong, T. D. Allston and K. J. Reed, J. Phys. Chem. C, 2009, 113, 15155–15159.
111. S. B. Putla, M. Kamali, B. Swapna, B. M. Reddy and P. Sudarsanam, ACS Appl. Nano Mater., 2024, 7, 6749–6771.
112. B. Liu, C. Li, G. Zhang, X. Yao, S. S. C. Chuang and Z. Li, ACS Catal., 2018, 8, 10446–10456.
113. F. Wang, S. Zhang, C. Li, J. Liu, S. He, Y. Zhao, H. Yan, M. Wei, D. G. Evans and X. Duan, RSC Adv., 2014, 4, 10834–10840.
114. Z. Jiang, D. Ding, L. Wang, Y. Zhang and L. Zan, Catal. Sci. Technol., 2017, 7, 3065–3072.
115. R. Si and M. Flytzani-Stephanopoulos, Angew. Chem., Int. Ed., 2008, 47, 2884–2887.
116. W. Wang, W. Xu, Z. Zhao, M. Cheng, M. Xun and H. Liu, Adv. Eng. Mater., 2024, 26, 2400092–2400111.
117. W. Gao, Z. Zhang, J. Li, Y. Ma and Y. Qu, Nanoscale, 2015, 7, 11686–11691.
118. J. Yang, H. Ding, Z. Zhu, Q. Wang, J. Wang, J. Xu and X. Chang, Appl. Surf. Sci., 2018, 454, 173–180.
119. E. Krumov, N. Starbov, K. Starbova, A. Perea and J. Solis, Appl. Surf. Sci., 2009, 256, 683–687.

---